Genetic differentiation that is exceptionally high and unexpectedly sensitive to geographic distance in the absence of gene flow: Insights from the genus Eranthis in East Asian regions

Abstract Genetic differentiation between populations is determined by various factors, including gene flow, selection, mutation, and genetic drift. Among these, gene flow is known to counter genetic differentiation. The genus Eranthis, an early flowering perennial herb, can serve as a good model to study genetic differentiation and gene flow due to its easily detectable population characteristics and known reproductive strategies, which can be associated with gene flow patterns. Eranthis populations are typically small and geographically separated from the others. Moreover, previous studies and our own observations suggest that seed and pollen dispersal between Eranthis populations is highly unlikely and therefore, currently, gene flow may not be probable in this genus. Based on these premises, we hypothesized that the genetic differentiation between the Eranthis populations would be significant, and that the genetic differentiation would not sensitively reflect geographic distance in the absence of gene flow. To test these hypotheses, genetic differentiation, genetic distance, isolation by distance, historical gene flow, and bottlenecks were analyzed in four species of this genus. Genetic differentiation was significantly high, and in many cases, extremely high. Moreover, genetic differentiation and geographic distance were positively correlated in most cases. We provide possible explanations for these observations. First, we suggest that the combination of the marker type used in our study (chloroplast microsatellites), genetic drift, and possibly selection might have resulted in the extremely high genetic differentiation observed herein. Additionally, we provide the possibility that genetic distance reflects geographic distance through historical gene flow, or adaptation in the absence of historical gene flow. Nevertheless, our explanations can be more rigorously examined and further refined through additional observations and various population genetic analyses. In particular, we suggest that other accessible populations of the genus Eranthis should be included in future studies to better characterize the intriguing population dynamics of this genus.


| INTRODUC TI ON
Assessment of genetic differentiation between populations allows for the characterization of taxa of interest and is therefore of great importance in the field of population genetics. Genetic differentiation, an indicator of how much populations are genetically isolated from others, is determined by various factors including gene flow, selection, mutation, and genetic drift (Slatkin, 1987). Among these, gene flow, which is the movement of genes between populations, is a critical factor that determines genetic differentiation (Ellstrand, 1992;Slatkin, 1987). Gene flow is generally known to counter genetic differentiation (Ehrlich & Raven, 1969;Rifkin et al., 2019;Slatkin, 1987). In plant populations, gene flow occurs either by pollen or seed dispersal (Ennos, 1994). Previous studies have argued that plant gene flow via wind and insect pollination, as well as by seed dispersal, can be spatially restricted (Ehrlich & Raven, 1969;Harper, 1977;Howe & Smallwood, 1982). For instance, in wind-pollinated Zea mays, only 1% of outcrossing contamination was observed in a population that was approximately 18 m from the source population (Ehrlich & Raven, 1969). As another example, in Clarkia and Delphinium, which are pollinated by insects, the distances shorter than 15m could be an effective barrier to the pollination (Roberts & Lewis, 1955). However, there are some cases for long-distance gene flow, for example, in tree species (Burczyk et al., 2004;Dow & Ashley, 1998;Liepelt et al., 2002;Ony et al., 2020;Schuster et al., 1989).
The genus Eranthis (Ranunculaceae), an early flowering perennial herb (Figure 1), provides an interesting case study for genetic differentiation and gene flow. This genus is composed of approximately 13 species, which are distributed from Europe to East Asia (Erst, Sukhorukov, et al., 2020;Erst et al., 2020b;Mabberley, 1993;Rukšāns, 2018;Tamura, 1990;Wang et al., 2001). This genus comprises relatively short plants (5-15 cm above ground) with tuberous rhizomes, palmately divided leaves and bracts, solitary flowers, and petaloid sepals (Oh & Ji, 2009;Sun et al., 1993). The life history and ecology of this genus have only been characterized in some species. For example, Eranthis hyemalis, which is distributed throughout Europe, is known to reproduce by vegetative reproduction using split tubers (Marcinkowski, 2002) and by sexual reproduction via insect pollination (Rysiak & Žuraw, 2011). Regarding growth conditions, E. byunsanensis in South Korea generally grows in gentle valleys with organic matter-rich soil, and this species may be sensitive to moisture conditions (Kim et al., 2012).
In the cases of E. byunsanensis, E. pungdoensis, and E. stellata, which inhabit South Korea, it was observed that there were only a small number of populations in a single province of South Korea, which means that the populations of these species are rarely observed. The populations of these species, whose sizes generally range from 300 to 20,000 m 2 , do not cluster together, and therefore become geographically isolated from each other (generally isolated by 10 km~, personal observation). In addition, it has been reported that populations are often separated by forest roads or mountain trails within the same site (Kim et al., 2012). The Eranthis species in other regions than South Korea also show similar distributional patterns (personal observation).
Meanwhile, some reports indicate that the seeds of the genus Eranthis cannot disperse over long distances, and from personal observation, the dispersal distance would be only up to 1 m. The species in this genus produce rounded or sub-globose seeds approximately 2-3 mm in diameter (Jung et al., 2010;Oh & Ji, 2009) which can only disperse through gravity and wind. Additionally, Thomson et al. (2011) found that seed dispersal distance is strongly correlated with plant height, suggesting that the seeds of the genus Eranthis (which is 5-15 cm tall above ground) may not disperse long distances.
Regarding insect pollination in this genus, even though insects carry pollen with relatively greater specificity than in the case of wind pollination, gene flow via this route is known to be more unlikely than originally expected (Ehrlich & Raven, 1969). For instance, bumblebees and anthophorid bees exhibit high visitation rates for target flowers, but travel only short distances between them (Rust, 1980;Kudoh and Whigham, 1997), meaning that insect-facilitated gene flow between populations is not likely to occur in some cases.
Furthermore, it has been suggested that early flowering in chilling temperatures may lead to low insect pollination in E. byunsanensis and Megaleranthis saniculifolia, the latter of which is morphologically and taxonomically similar to the genus Eranthis, resulting in low possibility of long-distance pollen dispersal (Jeong, 2010;Kim et al., 2012;Waser, 1982). Overall, it appears that gene flow via pollination

K E Y W O R D S
Eranthis, gene flow, genetic differentiation, geographic distance, population

T A X O N O M Y C L A S S I F I C A T I O N
Population genetics F I G U R E 1 Eranthis stellata of Gapyeong, Republic of Korea only occurs selectively in this genus. Given the observed characteristics of the populations for this genus, as well as previous literatures on gene flow, it seems very unlikely that gene flow is currently occurring between the populations of each species within this genus.
To date, only one study has characterized the population genetics of the genus Eranthis, which analyzed E. byunsanensis with 10 allozyme markers (So et al., 2012). In this study, considerably high genetic variation within population and little genetic differentiation among populations were observed. However, the restricted number of markers and the marker type used in the aforementioned study might have rendered less accurate conclusions regarding this species. Apart from this study, the population genetics of this genus has remained largely uncharacterized. In our study, we sought to investigate genetic differentiation and gene flow in the genus Eranthis, and tried to understand current and past population dynamics for this genus so that we can provide valuable fundamentals for population genetics in the genus Eranthis.
We hypothesized that due to the current low likelihood of gene flow between populations, as discussed above, the genetic differentiation between the populations in each species of this genus would be generally significant. Additionally, we hypothesized that genetic differentiation between the populations would not sensitively reflect the geographic distance between the populations in the absence of current gene flow.
To test these hypotheses, our study analyzed genetic differen-  (Oh & Ji, 2009;Sun et al., 1993). Due to their similar morphological and genetic characteristics, these two species are thought to be evolutionarily closely related (Oh & Oh, 2019). E. stellata is found in the Korean peninsula, the Jilin and Liaoning provinces of China, and some of the easternmost regions of Russia. E. pinnatifida ranges from the central to southern areas of Honshu, Japan. The morphological differences between these species are generally observed in the shapes of petals, leaves, and bracts.
Nectary patterns are also different between these species. The extracted DNA was then diluted to a final concentration of 10-50 ng/µl for the downstream analyses.
PCR was conducted in a TAKARA PCR Thermal Cycler Dice Touch (TAKARA Bio Inc.) using the Multiplex Master Mix of the QIAGEN Multiplex Master kit (QIAGEN). The PCR amplification conditions and procedures were the same as those described by Oh and Oh (2017).
An ABI3730xl DNA Analyzer (Applied Biosystems) and GeneMapper v. 3.7 (Applied Biosystems) were then used to measure PCR product lengths.
In the previous phylogeographic study of the authors (Oh & Oh, 2019), the same genotype data for 935 individuals of the genus Eranthis were used to analyze genetic diversity, population structure, and other evolutionary parameters.
We used both F ST and R ST because they both have their own drawbacks and we wanted these two measures to complement each other (Balloux and Lugon-Moulin, 2002). For example, F ST is known to be sensitive to mutation rates when migration is low (Balloux and Lugon-Moulin, 2002), and this can pose problems when analyzing microsatellite data, which shows high mutation rate. In contrast, R ST , which is suitable for analyzing markers with the stepwise mutation model (SMM), is less sensitive to high mutation rates reported in microsatellite, thus complementing the weakness of the F ST (Holsinger & Weir, 2009). Meanwhile, due to the high variance of R ST , F ST may outperform R ST even under strict SMM (Gaggiotti et al., 1999;Balloux and Lugon-Moulin, 2002). In fact, there is no mutation model that perfectly fits the behavior of microsatellites, and therefore F ST and R ST are often used together in microsatellite data analyses (Ando et al., 2011;Balloux and Lugon-Moulin, 2002;De March et al., 2002;Sander et al., 2018).

| Genetic distance
As an additional tool for understanding the genetic differences between the populations in this genus, genetic distance was measured so that this parameter can effectively complement the genetic differentiation analysis above. For each Eranthis species, genetic distances between populations were calculated using the Microsatellite Analyzer (MSA) 4.05 (Dieringer & Schlötterer, 2003).
Among the many genetic distance parameters available, (δµ) 2 , which was developed by Goldstein et al. (1995), was used for the analysis of microsatellite data. (δµ) 2 was developed based on stepwise mutation model which can be successfully applied to microsatellite loci (Goldstein et al., 1995). (δµ) 2 is known to be a linear function of the time since divergence between the populations (Goldstein et al., 1995). The values for (δµ) 2 in this study provide an estimation of the evolutionary distances between the populations.

| Isolation by distance
Our study sought to determine if there were significant relationships between the genetic distances and geographic distances for each species. The Mantel test (Mantel, 1967), which is a broadly used statistical test to evaluate the association between distance matrices, was conducted to assess the correlation between genetic distance (F ST in this study) and geographic distance. In other words, this test was conducted to identify whether we can find the pattern of isolation by distance in the data. The Mantel tests were conducted using GenAlEX 6.5 (Peakall & Smouse, 2012).

| Gene flow
The coalescence-based MIGRATE-N 3.6.11 (Beerli & Palczewski, 2010) was used to infer the level and pattern of historical gene flow in each species. In the case of E. stellata, the 20 populations that were analyzed together in the above analyses were divided into   (Oh & Oh, 2019)  were extremely high (i.e., higher than 50; Table S1). This result was attributed to the genetic differences between the SCP population and the rest of the E. stellata populations, which was elucidated in a previous study (Oh & Oh, 2019), where the reason for this difference was not identified.

| Isolation by distance
Each of the three groups identified above was analyzed to determine whether there was a clear relationship between genetic distance (in this analysis, F ST ) and geographic distance in the focal species, and the Mantel test (Figure 3) was used here. The Mantel test identified a negative correlation in the E. byunsanensis and E. pungdoensis populations, which indicates that isolation by distance is not established in this group (R 2 = .1347, p = .16; Figure 3a). In E. pinnatifida, we observed a positive correlation between genetic distance and geographic distance, which indicates the existence of isolation by distance in these populations (R 2 = .3464, p = .05; Figure 3b). In the case of E. stellata, genetic distance increased with geographic distance, as observed in E. pinnatifida, and isolation by distance was also identified in these populations (R 2 = .1395, p = .01; Figure 3c).
Taken together, our Mantel test results show that isolation by distance is generally established in this genus.

| Historical gene flow
The long-term effective population size (θ), the long-term migra-  (Table S2). In these populations, the effective population sizes were comparable, ranging from 0.25914 to 0.32588 (Table S3).
In E. pinnatifida, the migration rates ranged from 4.126 (95% CI: 0.0-11.5) to 79.766 (95% CI: 56.5-98.0), and these values were much higher than those of the other groups (Table S2). The migrant numbers were also very high, with the highest value being 270.0269 (Table S2). In this species, the effective population sizes ranged from 0.57246 to 4.96072 (Table S3).
The relationship between migration rate, which was derived from historical gene flow analyses, and geographic distance was investigated as well to obtain additional insights into the gene flow dynamics of this genus. Except for the Chinese E. stellata populations (Figure 4d), the four remaining groups exhibited negative correlations between migration rate and geographic distance (Figure 4a-c,e). This indicates that there existed a notable relationship between the degree of migration and the geographic distance in the focal species and that geographic distance markedly constrained successful migrations in these populations.

| Bottleneck analysis
According to the BOTTLENECK test, a large number of the popula- In a study by So et al. (2012), the genetic differentiation between five populations of E. byunsanensis was estimated using 10 allozyme markers, and the mean F ST value was 0.131. This inconsistency between F ST values in the So et al. study and our study can be partially explained by the genetic markers used in the analyses. Microsatellites are generally known to evolve much faster than allozyme loci, and better detect genetic differentiation between populations, with mutation rates of 10 −3 to 10 −4 (Dallas, 1992;Weber & Wong, 1993) compared to the 10 −6 to 10 −9 of the allozyme loci (Ayala, 1976). In previous studies on red pine (Pinus resinosa), no allozyme diversity was detected, whereas 23 haplotypes were recovered with 9 chloroplast microsatellite markers (Echt et al., 1998;Provan et al., 2001). Therefore, the high population differentiation values in our study likely derived from the high mutation rates of chloroplast microsatellites.
Genetic drift might have also influenced the high levels of genetic differentiation. In extreme cases, genetic drift can cause allele fixation in opposite directions in many populations, resulting in very high genetic differences between the populations (Jeong, 2010;Nistelberger et al., 2015). In smaller or isolated populations, genetic drift has an even stronger effect (Toczydlowski & Waller, 2019). In this study, the sampled populations were isolated and generally small, with only   (Caballero & Hill, 1992;Lowe et al., 2005). Previous studies on E. byunsanensis (So et al., 2012) and E. stellata (Jeong et al., 2005) suggested that inbreeding is significantly occurring in these populations, which may ultimately contribute to genetic differentiation between the populations. However, additional studies on inbreeding in the genus Eranthis should be conducted to reach more reliable conclusions.
Finally, the high genetic differentiation observed herein might have also been due to selection. Selection can genetically structure populations when organisms are selectively adapted to gradient environments, which is also referred to as isolation by environment or isolation by adaptation (Nosil et al., 2008;Sexton et al., 2014;Wang & Summers, 2010). For example, the E. byunsanensis and E. pungdoensis populations are geographically separated from the others (>150 km), and different selective pressures likely acted in the different locations where these populations were located, resulting in different genetic characteristics between the populations. Even though morphological differences are not clearly observed between these widely distributed populations, adaptive genotypes might have still been generated, resulting in high genetic differentiation. However, in our study, selection as an important factor affecting genetic differentiation is not yet supported by actual analyses or observations, and further research is needed to define the extent to which the high genetic differentiation observed is due to selection. Overall, the high genetic differentiation values in our study are likely due to a combination of all of the aforementioned factors, which are the genetic marker used, genetic drift, and possibly selection.

| Unexpected correlation between genetic distance and geographic distance in the genus Eranthis
As mentioned above, we initially hypothesized that geographic distances between populations would not meaningfully affect genetic distances in the absence of gene flow considering that gene flow would be the major factor which can generate the correlation be-   Overall, as discussed above, genetic differentiation and genetic distance appeared to reflect the geographic distance in many cases within the genus Eranthis. However, additional studies are required to identify the mechanisms by which genetic differentiation reflects geographic distance in Eranthis populations in cases where gene flow between populations seems to be impossible. Here, we propose two different hypotheses that could explain the counterintuitive correlations between geographic distance and genetic distance (genetic differentiation) observed herein.
First, it is probable that gene flow historically existed, and this past event determined the current correlation between genetic differentiation and geographic distance. Our historical gene flow analysis using MIGRATE-N demonstrates that there existed substantial gene flow long time ago, supporting this hypothesis. The occurrence of past gene flow means that geographic distance had previously affected gene flow (the correlation between migration rate and geographic distance was verified in our study; Figure  and other abiotic / biotic conditions. Generally, populations that are separated by long distances tend to inhabit substantially different environments and experience different evolutionary pressures compared to populations in closer proximity. This correlation between geographic distance and environmental difference is well established (Lee & Mitchell-Olds, 2011;Wang & Bradburd, 2014). Therefore, the distance between the populations might have affected genetic differentiation via adaptation. Additionally, it is also likely that different mutations occurred with time within each population, thus further contributing to the genetic differences between the populations (Slatkin, 1987). These two hypotheses proposed herein should be further refined and tested to explain the intriguing and counterintuitive population genetic phenomena observed in this genus.

| CON CLUS ION
In this study, we observed interesting phenomena regarding genetic differentiation and gene flow in the genus Eranthis, and provided preliminary insights into the population genetics and dynamics of this genus. The extremely high genetic differentiation observed herein is likely due to a combination of the genetic markers used to conduct our analyses, genetic drift, and possibly selection. Unexpected positive correlations were identified between genetic distance and geographic distance in the absence of gene flow and two hypotheses were proposed to explain these observations. The current correlation between genetic distance and geographic distance may have resulted from past gene flow, or the action of adaptation to different environments in different locations. It is also probable that both phenomena acted in combination. We recognize that there can be further reasonable explanations for our observations and that future studies should incorporate all of the existing populations of these species to better characterize their population dynamics. In conclusion, our research provides an important case study that addressed intriguing and fundamental population genetic questions, and establishes a precedence for the analysis of population dynamics in the genus Eranthis, a lesser-studied wild plant. In addition, our findings indicate that this genus can provide important and unique insights into the mechanisms of genetic differentiation and gene flow, and could therefore serve as a valuable system to facilitate the understanding of unexplored population genetics phenomena.

ACK N OWLED G EM ENT
This work was supported by the National Research Foundation of Korea (NRF-2015R1D1A1A01057391).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data presented in this study has been uploaded to Dryad (https://doi.org/10.5061/dryad.w9ghx 3frr).